Analysis and correction of off‐resonance artifacts in echo‐planar cardiac diffusion tensor imaging

Purpose Cardiac diffusion tensor imaging using EPI readout is prone to image distortions in the presence of field inhomogeneities. In this work, a framework to analyze and correct image distortions in cardiac diffusion tensor imaging is presented. Methods A multi‐coil reconstruction framework was implemented to enable field map‐based off‐resonance correction. Numerical simulations were used to examine reconstruction performance for EPI phase‐encode directions blip up‐down and down‐up for different degrees of off‐resonance gradients and varying field map resolution. The impact of coil encoding was analyzed using the g‐factor and normalized RMSE. Finally, the proposed method was tested on free‐breathing in vivo cardiac diffusion tensor imaging data acquired in healthy subjects at 3 Tesla. Results Depending on the local field map gradient strength and polarity and the selected phase‐encode direction, field inhomogeneities lead to either local spatial compression or stretching with standard image reconstruction. Although spatial compression results in loss of image resolution upon field map‐based reconstruction, spatial stretching can be recovered once multiple receive coils are utilized. Multi‐coil reconstruction was found to reduce the normalized RMSE from 34.3% to 8.1% for image compression, and 33.6% to 1.8% for image stretching, with resulting average g‐factors 14.7 ± 2.9 and 1.2 ± 0.1, respectively. In vivo, multi‐coil field map‐based reconstruction yielded improved alignment of angle maps with anatomical cine data. Conclusion Multi‐coil, field map‐based image reconstruction for echo‐planar cardiac diffusion tensor imaging allows accurate image reconstruction provided that the phase‐encode direction and polarity is chosen to principally align with the direction and polarity of the prominent gradients of field inhomogeneities.